CN101809653A - A method and an apparatus for processing an audio signal - Google Patents

Abstract

A method for processing an audio signal, comprising: receiving the audio signal; and processing the received audio signal, wherein the audio signal is processed according to a scheme comprising: comparing a size information of at least two blocks of A+1 level with a size information of a block of A level corresponding to the at least two of A+1 level; and, determining the at least two blocks of A+1 level as an optimum block if the size information of the at least two blocks of A+1 level is less than the size information of the block of A level is disclosed. A method for processing an audio signal, comprising: receiving the audio signal; and processing the received audio signal, wherein the audio signal is processed according to a scheme comprising: comparing a size information of a block of A level with a size information of at least two blocks of A+1 level; and, determining the block of A level as an optimum block if the size information of the block of A level is less than the size information of the at least two blocks of A+1 level is disclosed.

Description

The method and apparatus that is used for audio signal

Technical field

The present invention relates to be used for the method and apparatus of audio signal, more specifically, relate to the method and apparatus that is used for coding audio signal.

Background technology

Past has been realized the storage and the playback of sound signal by many distinct methods.For example, by phonograph technology (player of for example recording), magnetic technology (for example magnetic tape cassette) and digital technology (for example CD) record and preservation music and voice.Along with the audio storage technical development, need overcome many challenges to optimize the quality and the storability of sound signal.

At the file and the wideband transmit of music signal, harmless reproduction is becoming feature more even more important than high-level efficiency in by the compression of perception, has the demand of open and general compression scheme in content holder and broadcasting equipment.Respond this demand, considered to adopt new lossless coding scheme.Lossless audio coding is owing to the perfect reproduction of original signal allows without any mass loss ground compression digital audio frequency data.

Disclosure

Technical matters

Yet in the lossless audio coding method, coding costs a lot of money the time, needs ample resources, and has very high complexity.

Technical scheme

Therefore, the present invention is directed to the method and apparatus that is used for audio signal of the one or more problems that cause from limitation and the defective avoided in essence by corresponding technology.

An object of the present invention is to provide a kind of method and apparatus that is used for lossless audio coding, described method and apparatus is owing to the perfect reproduction of original signal allows to compress without any the digital audio-frequency data of mass loss.

Another object of the present invention provides a kind of method and apparatus that is used for lossless audio coding, thereby reduces scramble time, computational resource and complexity.

The part of other advantage of the present invention, purpose and feature will be illustrated in the following description, bright Liao or can know from the practice of the present invention and another part will become after the content below those skilled in that art read carefully.Purpose of the present invention and other advantage will be realized and reached by the structure of specifically noting in written explanation and claims and the accompanying drawing.

Beneficial effect

The invention provides following effect and advantage.

At first, the present invention can provide a kind of method and apparatus of lossless audio coding that is used for to reduce scramble time, computational resource and complexity.

Secondly, the present invention can accelerate the piece handoff procedure in the audio frequency lossless coding.

Moreover the present invention can reduce complexity and the computational resource in the long-term forecasting process of audio frequency lossless coding.

Description of drawings

Comprise providing the accompanying drawing that the present invention is further understood and constitute the application's a part that embodiments of the invention are shown, and be used for explaining principle of the present invention with instructions.In the accompanying drawings:

Fig. 1 is the exemplary explanation according to scrambler 1 of the present invention.

Fig. 2 is the exemplary explanation according to demoder 3 of the present invention.

Fig. 3 is the exemplary explanation according to the bitstream structure of the compressed sound signal that comprises a plurality of passages (for example M passage) of the present invention.

Fig. 4 is the schematic block diagram according to the piece switching device shifter that is used for audio signal of first embodiment of the invention.

Fig. 5 is the exemplary explanation according to the theory figure of hierarchical block dividing method of the present invention.

Fig. 6 is the exemplary explanation according to the of the present invention variable combination of cutting apart.

Fig. 7 is an exemplary plot of explaining the theory of the block switching method that is used for audio signal according to an embodiment of the invention.

Fig. 8 is the exemplary process diagram that is used for the block switching method of audio signal according to an embodiment of the invention.

Fig. 9 is an exemplary plot of explaining the theory of the method that is used for audio signal according to another embodiment of the present invention.

Figure 10 is the example flow diagram that is used for the block switching method of audio signal according to another embodiment of the present invention.

Figure 11 is the exemplary process diagram according to the block switching method that is used for audio signal of another embodiment of the present invention version.

Figure 12 is an exemplary plot of explaining Figure 11 theory.

Figure 13 is the example block diagram according to the long-term forecasting device that is used for audio signal of the embodiment of the invention.

Figure 14 is the exemplary process diagram according to the long-range forecast method that is used for audio signal of the embodiment of the invention.

Best mode

In order to obtain according to these purposes that the invention is intended to and other advantage, such as herein embodiment and the broad description, a kind of method of audio signal comprises: received audio signal; And handle the sound signal received, wherein according to a kind of scheme audio signal, described scheme comprises: the size information of at least two pieces of A+1 layer and size information corresponding to the A layer piece of at least two pieces of described A+1 layer are compared; And if the size information of at least two pieces of described A+1 layer is less than the size information of A layer piece, then at least two pieces with described A+1 layer are defined as optimical block, and wherein sound signal may be partitioned into the piece of several layers with hierarchy.

In another aspect of this invention, the method that is used for audio signal comprises: received audio signal; And handle the sound signal received, wherein according to a kind of scheme audio signal, described scheme comprises: the size information of the A layer piece of the size information of at least two pieces of A+1 layer and sound signal entire frame is compared; And if whole size information of at least two pieces of described A+1 layer less than be included in this frame in the size information of the corresponding A layer piece of at least two pieces of A+1 layer, then at least two pieces with described A+1 layer are defined as optimical block.

In another aspect of this invention, a kind of method that is used for audio signal comprises: received audio signal; And handle the sound signal received, and wherein handle described sound signal according to a kind of scheme, described scheme comprises: the size information of at least two pieces of the size information of A layer piece and A+1 layer is compared; The size information of at least two pieces of the size information of A+1 layer piece and A+2 layer is compared; And if the size information of A layer piece is less than the size information of at least four pieces of the size information of at least two pieces of A+1 layer and A+2 layer, then the piece with the A layer is defined as optimical block.

In another aspect of this invention, a kind of method that is used for audio signal comprises: received audio signal; Handle the sound signal that is received; Wherein according to a kind of scheme audio signal, described scheme comprises: the size information of at least two pieces of the size information of A layer piece and A+1 layer is compared; And if the size information of A layer piece then is defined as optimical block with A layer piece less than the size information of at least two pieces of A+1 layer.

In another aspect of this invention, a kind of method that is used for audio signal comprises: received audio signal; And handle the sound signal received; Wherein according to a kind of scheme audio signal, described scheme comprises: the size information of A layer piece and size information corresponding at least two pieces of the A+1 layer of the A layer piece of sound signal entire frame are compared; And if whole size information of A layer piece then are defined as optimical block with A layer piece less than the size information corresponding at least two pieces of the A+1 layer that is included in the A layer piece in the frame.

In another aspect of this invention, a kind of device that is used for audio signal, comprise: initial comparing unit, described initial comparing unit compares the size information of at least two pieces of A+1 layer and size information corresponding to the A layer piece of at least two pieces of A+1 layer; And condition comparing unit, if the size information of at least two pieces of A+1 layer is less than the size information of A layer piece then described condition comparing unit is defined as optimical block with at least two pieces of described A+1 layer, wherein sound signal may be partitioned into have several layers piece to form hierarchy.

In another aspect of this invention, a kind of device that is used for audio signal comprises: received audio signal; And handle the sound signal received; Wherein according to a kind of scheme audio signal, described device comprises: initial comparing unit, and described initial comparing unit compares the size information of at least two pieces of the size information of A layer piece and A+1 layer; And the condition comparing unit, if the size information of A layer piece is less than the size information of at least two pieces of A+1 layer then described condition comparing unit is defined as optimical block with A layer piece.In another aspect of this invention, a kind of method that is used for audio signal comprises: received audio signal; And handle the sound signal received; Wherein according to a kind of scheme audio signal, described scheme comprises: the size information of at least two pieces of A+1 layer and size information corresponding to the A layer piece of at least two piece of A+1 layer are compared; If the size information of at least two pieces of A+1 layer is less than the size information of A layer piece, then at least two pieces with the A+1 layer are defined as optimical block; And determine lag information based on the auto-correlation function value of the sound signal that comprises optimical block; And based on lag information estimation long-term forecasting filter information.In another aspect of this invention, the device that is used for audio signal comprises: initial comparing unit, and described initial comparing unit compares the size information of at least two pieces of A+1 layer and size information corresponding to the A layer piece of at least two pieces of A+1 layer; The condition comparing unit is if the size information of at least two pieces of A+1 layer is less than the size information of A layer piece then described condition comparing unit is defined as optimical block with at least two pieces of A+1 layer; Lag information is determined parts, and described lag information determines that parts determine lag information based on the auto-correlation function value of the sound signal that comprises optimical block; And the filter information estimation section, described filter information estimation section is estimated the long-term forecasting filter information based on lag information.

The general remark and the ensuing detailed description that are appreciated that front of the present invention are exemplary and illustrative, and aim to provide the present invention such as claims are described further specifies.

The invention pattern

Now in detail referring to preferred embodiment of the present invention, the example shown in the drawings.As far as possible, all representing same or analogous parts with identical Reference numeral in the accompanying drawing.

Before the beginning description of the invention, note most term disclosed by the invention corresponding to known general terms in the industry, but the applicant selects some terms as required, and open in the description below the present invention.Therefore, the term of understanding by applicant's definition based on the implication that the present invention gave is preferable.

In the lossless audio coding method,, can the determinacy mode realize both some parts of encoder owing to cataloged procedure is the data degradation that do not have of completely reversibility.

The structure of codec

Fig. 1 is the exemplary plot according to scrambler 1 of the present invention.Referring to Fig. 1, piece switching part 110 can be configured to the sound signal of input is cut apart framing.Can be used as broadcasting or the sound signal of this input of reception on digital media.In a frame, a plurality of passages can be arranged.Each passage can further be divided into the audio sample piece for further processing.

The frame sampling that impact damper 120 can be configured to storage block and/or cuts apart by piece switching part 110.Coefficient estimation parts 130 can be configured to estimate the best set of the coefficient value of each piece.Can select the number of coefficient, i.e. time item of fallout predictor with adapting to.In operation, 130 pairs of digital audio blocks of coefficient estimation parts calculate one group of PARCOR (partial auto correlation) (hereinafter being called " PARCOR ") value.The PARCOR of PARCOR value indication predictor coefficient characterizes.Afterwards, can dispose and quantize parts 140, quantize with group to the PARCOR value obtained from coefficient estimation parts 130.

The first entropy coding parts 150 can be configured by from the PARCOR value and deduct bias and calculate the PARCOR residual value, and use by the entropy code of entropy parameter definition the PARCOR residual value is encoded.Here, optimize table from one and select bias and entropy parameter choosing, select this optimization table from a plurality of tables based on the sampling rate of DAB piece.Can be at the described a plurality of tables of a plurality of sampling rate scope predefines for the optimization compression digital audio frequency data of transmitting.

Coefficient converting member 160 can be configured to convert the PARCOR value through quantizing to linear predictive coding (LPC) coefficient.In addition, short-term forecasting device 170 can be configured to use the before crude sampling of linear forecast coding coefficient from be stored in impact damper 120 to estimate current predicted value.

In addition, first subtracter 180 can be configured to use original value that is stored in the digital audio-frequency data in the impact damper 120 and the predicted value of estimating in short-term forecasting device 170 to calculate the prediction residual value of DAB piece.Long-term predictor 190 can be configured to estimate lag information τ and LTP filter information γ _j, and set the flag information that indicates whether to carry out long-term forecasting, and use lag information and LTP filter information to produce long-term forecasting value e^ (n).

Configurable second subtracter 200 with after long-term forecasting, uses current predicted value e (n) and long-term forecasting value e^ (n) to estimate new residual value e～(n).In Figure 13 and Figure 14, set forth the details of the long-term predictor 190 and second subtracter 200.

The second entropy coding parts 210 are configured to use different entropy codes to prediction residual value coding and generation code index.The index of selected code is as secondary (or auxiliary) information transmission.

The second entropy coding parts 210 of prediction residual value provide two kinds of optional coding techniquess with different complexities.A kind of Golomb-Rice of being coding (abbreviating " Rice code " hereinafter as) method, and another kind is Block Gilbert-Moore code (abbreviating " BGMC " hereinafter an as) method.Except the low but still high efficiency Rice code of complicacy, BGMC arithmetic coding scheme is that cost provides much better compressibility slightly to increase complexity.

At last, multiplexing components 220 can be configured to carry out multiplexed to form the bit stream of compression to encoded prediction residual value, code index, encoded PARCOR residual value and other additional information.Scrambler 1 also provides the Cyclic Redundancy Check inspection, provides it mainly in order to make the data of demoder verification through decoding.In coder side, but CRC can be used to guarantee that compressed data are losslessly encodings.In other words, CRC can be used to compressed data lossless ground decoding.

The additional code option comprises flexible piece handover scheme, random access and joint channel coding.Scrambler 1 can use in these options arbitrarily, and item provides some hierarchy compressions with different complexities.The joint channel coding is used for utilizing the dependency between stereo channel or the multi channel signals.This can obtain by the difference between two passages in the section is encoded, and more effectively this species diversity is encoded for one in the wherein comparable Src Chan.

Fig. 2 is the exemplary plot according to demoder 3 of the present invention.More particularly, Fig. 2 illustrates owing to do not carry out and adapt to and complexity is not starkly lower than the lossless audio decoding signals of scrambler.

Multichannel Knock-Down Component 310 can be configured by broadcasting or received audio signal and the coded prediction residual value of DAB piece, code index, encoded PARCOR residual value and other additional information multichannel decomposed on digital media.

First entropy decoding parts 320 can be configured to use the entropy code by the entropy parameter definition that the PARCOR residual value is decoded, and by adding that bias calculates one group of PARCOR value through the PARCOR residual value of decoding.Here, bias and entropy parameter are selected from a table, select this table based on the sampling rate of DAB piece by scrambler from a plurality of tables.

Second entropy decoding parts 330 can be configured to use code index that the coded prediction residual value of decomposing through multichannel is decoded.Long-term predictor 340 can be configured to use lag information and LPT filter information to estimate long-term predictor.In addition, first adder 350 can be configured to use long-term forecasting value e^ (n) and residual value e～(n) to calculate short-term LPC residual value e (n).

Coefficient converting member 360 can be configured to convert the PARCOR value of entropy decoding to the LPC coefficient.In addition, short-term forecasting device 370 can be configured to use the prediction residual value of LPC coefficient estimation DAB piece.380 of second adders are configured to the prediction of using short-term LPC residual value e (n) and short-term forecasting device to calculate digital audio-frequency data.At last, assembling parts 390 is configured to the piece through decoding is assembled into frame data.

As discussed above, demoder 3 can be configured to encoded prediction residual value and PARCOR residual value are decoded, and converts the PARCOR residual value to the LPC coefficient, and adopts the inverse prediction wave filter to calculate harmless reproducing signal.The computing power of demoder 3 depends on scrambler 1 selected prediction time item.In most cases, even real-time decoding also is possible in end systems.

Fig. 3 is the exemplary plot according to the bitstream structure of the compressed sound signal that comprises a plurality of passages (for example M passage) of the present invention.

Bit stream is made of at least one audio frame, and this audio frame comprises a plurality of passages (for example M passage).Use is divided into a plurality of according to of the present invention handover scheme with each passage, and this will describe in detail hereinafter.Each piece of cutting apart has different sizes and comprises coded data according to Fig. 1.For example, the coded data in the piece through cutting apart comprises the residual value of code index, a prediction time K, predictor coefficient and coding.If in pairs adopting combined coding between the passage, then to cut apart be identical to the piece of two passages, and with the interleaving mode storage block.Otherwise it is independently that the piece of each passage is cut apart.

Hereinafter, will be described in detail data switching and long-term forecasting in conjunction with following accompanying drawing.

Piece switches

Fig. 4 is the example block diagram according to the piece switching device shifter that is used for audio signal of the embodiment of the invention.As shown in Figure 4, the device that is used for processing audio comprises piece switching part 110 and impact damper 120.More specifically, partition member 110 comprises partition member 110a, initial comparing unit 110b and condition comparing unit 110c.Partition member 110a is configured to each passage of frame is divided into a plurality of, and identical with the switching part of mentioning in conjunction with Fig. 1 before 110.In addition, be used to store the impact damper of cutting apart by the piece of piece switching part 110 selections 120 and be similar to the impact damper of mentioning in conjunction with Fig. 1 before 120.

Details and the process of partition member 110a, initial comparing unit 110b and condition comparing unit 110c is called as " end of to the top method " and/or " pushing up certainly to end method ".

At first, partition member 110a can be configured to each passage layeredly is divided into a plurality of inferiorly.Fig. 5 is the exemplary explanation according to the theory figure of the dividing method of piece by different level of the present invention.

Fig. 5 illustrates the method that a frame is divided into by different level 2 to 32 pieces (for example 2,4,8,16 and 32).When providing a plurality of passage in a frame, each passage can be divided into (or being divided into) nearly 32 pieces.As shown in the figure, each passage constitutes a frame through block.For example, for layer=5, a frame is divided into 32 pieces.In addition, as previously mentioned, can in module unit, carry out prediction and entropy coding through cutting apart.

Fig. 6 is the exemplary plot that illustrates according to the multiple combination of block of the present invention.As shown in Figure 6, can in a frame, form the N of cutting apart of piece combination in any _B=N, N/2, N/4, N/8, N/16 and N/32, if each piece be by Double Length high-level block cut apart formation.That is, top block length equals 32 times of bottom block length.

For example, shown in the example of Fig. 5, a frame can be divided into N/4+N/4+N/2, yet a frame can't be divided into N/4+N/2+N/4 ((e) for example shown in Figure 6 and (f)).The piece dividing method is associated with the process of selecting suitable piece to cut apart.Hereinafter, be called as " end of to the top method " and " top certainly " according to of the present invention dividing method to end method.

The end of to the top method

Fig. 7 is the exemplary plot of the theory of the block switching method that is used for audio signal of the explanation according to the embodiment of the invention.Fig. 8 is the exemplary process diagram according to the block switching method that is used for audio signal of the embodiment of the invention.

Referring to Fig. 7, in six layers each, a=0 ... 5, the audio frame that N is sampled is divided into length N _B=N/B=N/2 ^aB=2 ^aIndividual piece.Here, think that a layer a=0 is top or top layer, and think that a layer a=5 is lowermost layer or bottom.In addition, for the end of to the top method, first corresponding to lowermost layer, second corresponding to the next higher level (a=4) on the lowermost layer, the 3rd corresponding to the next higher level (a=3) on second, by that analogy.In some cases, first, second and the 3rd can be applicable to layer a=4 to the layer a=2, the layer a=3 to the layer a=1 or the layer a=2 to the layer a=0 piece.

All pieces to a layer (or in one deck) are all encoded, and encoded piece is stored together by its big or small S (is unit with the position) separately temporarily.Size S is corresponding to one in coding result, position size and the coded data piece.To each the layer encode, thereby to each the piece formation value S in each layer (a, b), b=0 ... B-1.In some cases, do not need the piece that will skip is encoded.

Then, from lowermost layer a=5, two adjacent pieces and at least one piece of higher level a=4 can be compared.That is to say, with the position size of the position size of two adjacent blocks of layer a=5 and relevant block relatively to determine which piece needs position size still less.Here, relevant block refers to the block size with regard to length/duration of cutting apart.For example, initial two adjacent blocks (from the left side) of lowermost layer a=5 are corresponding to the initial piece (from the left side) of the second lowermost layer a=4.

Referring to Fig. 4 and Fig. 8, initial comparing unit 110b compares (S110) with the position size of two first (bottom) and second position size.Two first position size can equal one first and another big or small sum of first.At bottom is under the situation of a=5, and the comparison sheet among the step S110 is shown following equation 1.

Equation 1

S(5，2b)+S(5，2b+1)≥S(4，b)

If two first position size is less than second position size (being judged to be "No" among the step S110), then initial comparing unit 110b selects two first (S120) of lowermost layer.In other words, this two first is stored in the impact damper 120, and second is not stored in the impact damper 120, and in step S120 it is deleted from the impact damper of odd-job, this is not have raising owing to compare second with regard to bit rate.Behind step S120, stop comparison and selection, and the relevant block of following one deck is no longer carried out comparison and selection.

Perhaps, if the big or small position size (being judged to be "Yes" among the step S110) that is equal to or greater than second in two first position, then condition comparing unit 110c compares (S130) with two second position size and the 3rd position size.In some cases, in step S110, if at least one in two first the position size less than second position size, execution in step S130 then, described second all piece (b=0 with a layer ... B) two first correspondences in.The condition of this correction can be applicable to following step S150 and S170.If two second position size is less than the 3rd position size (being judged to be "No" among the step S130), then condition comparing unit 110c selects two second (S140).In step S140, the long piece of layer 4 has replaced two short blocks from layer 5.Behind step S140, compare and selection processing termination.

Similar with step S130 and S140, the 3rd and the 4th the comparison (S150) of layer a=2 of execution level a=3, and the result carries out selection (S160) based on the comparison.In general, condition comparing unit 110c only just compares (S170) with the position size of two i pieces and the position size of i+1 piece when the position size of two i pieces (layer a=a+1) is equal to or greater than the position size of i+1 piece (layer a=a), and compares (S180) according to the suitable piece of comparative result selection or to following one deck.Step S170 is expressed as following equation 2.But repeating step S170 is till arriving top (a=0).

Equation 2

S(a+1，2b)+S(a+1，2b+1)≥S(a，b)

A=0 wherein ... 5, b=0 ... B-1,

" a+1 " corresponding to the layer of i piece, " a " is corresponding to the layer of i+1 piece.

Refer again to Fig. 7, illustrate with lead and may be selected to be the piece that share piece, the piece that can't benefit is shown from further merging, and the piece that must handle is shown with white with bright grey.In addition, illustrate unwanted or obsolete with grey (or translucent), this expression can be saved comparison procedure.There is not progress from layer a=3 to layer a=1, therefore do not need to handle higher level a=1 and a=0.At last, selecting the piece of layer a=3 is b=0 ... 7, selecting the piece of layer a=4 is b=8 ... 15 ..., selecting the piece of layer a=5 is b=20-21, saves remaining.

Can come performing step S110 to step S180 by following C class false code 1, but this be construed as limiting to the present invention.Specifically, realize false code 1 according to correction conditions above-mentioned.

False code 1

for(a＝5；a＜＝0；a-){//for?all?levels

B＝1＜＜a； //block?length?in?level?a

for(b＝0；b＜B；b++){//fbr?all?blocks

size[a][b]＝EncodeBlock(x+b＊B，buf[a][b])；//encode?block?and?store?in?buf

}

if(a＜5){ //if?not?lowest?level

improved＝0；

for(b＝0；b＜B；b++){//compare?size?of?current?block?with?size?of?two?blocks?in

level?a+1

if(size[a][b]＞size[a+1][2*b]+size[a+1][2＊b+1]){//copy?two?short?blocks?from

level?a+1?into?the?long?block?of?level?a

memcpy(buf[a][b]，buf[a+1][2＊b]，size[a+1][2＊b])；

memcpy(buf[a][b]+size[a+1][2＊b]，buf[a+1][2＊b+1]，size[a+1][2＊b+1])；

size[a][b]＝size[a+1][2＊b]+size[a+1][2＊b+1]； //update?size?of?new

longblock

}

else

improved＝1；//improvement?by?longer?blocks

}

if(limproved)

break；//stop?iteration?atlevel?a

}

}

From pushing up to end method

Fig. 9 is an exemplary plot of explaining the theory of the block switching method that is used for audio signal according to another embodiment of the present invention.Figure 10 is the example flow diagram that is used for the block switching method of audio signal according to another embodiment of the present invention.Referring to Fig. 9, with identical to the top method, for six layer a=0 the end of from ... each layer of 5 is divided into length N with N audio frame of sampling _B=N/B=N/2 ^aB=2 ^aIndividual piece.To the top method, concerning pushing up certainly to the end method, first corresponding to top (a=0) end of from contrast, and second corresponding to top one deck (a=1) down, and the 3rd corresponding to following one deck (a=2) of second, but this is not construed as limiting the present invention.In some cases, first, second and the 3rd can be applicable to layer a=1 to the layer a=3, the layer a=2 to the layer a=4 or the layer a=3 to the layer a=5 piece.

Be similar to the end of to the top method to end method from the top, promptly do not cause progressive some place to end search, except locating beginning at top layer (a=0) and towards lower level carries out at one deck down.Locate at each layer " a ", the size of a piece and two relevant block of following one deck a+1 are compared.If these two short blocks need figure place still less, then replace (being virtual dividing) long layer " a " piece, and algorithm advances to a layer a+1.Otherwise if long piece needs less position, then adaptation terminates in lower layer.

Referring to Fig. 4 and Figure 10, initial comparing unit 110b compares (S210) with the position size of (top layer) first and two second position size.Size and another big or small sum of second that the big I in two second position equals one second.At top layer is under the situation of a=0, and the comparison sheet among the step S210 is shown following equation 3.

Equation 3

S(0，b/2)≥S(1，b)+S(1，b+1)

Identical with the step S120 of front, if first position size is less than two second big or small (being judged to be "No") among the step S110, then initial comparing unit 110b selects two top first (S220).Otherwise if the big or small position size (being judged to be "Yes" among the step S110) that is equal to or greater than two second in first position, then condition comparing unit 110c compares (S230) with second position size and two the 3rd position size.In some cases, at step S210, if the position size of at least one in first less than with whole piece (b=0 of a layer ... two second position size of first correspondence B), then execution in step S230.This correction conditions is applicable to following step S250 and S270.Identical to step S180 with step S140, execution in step S240 is to step S280.Expression step S270 is following equation 4.But repeating step S270 is till arriving lowermost layer (a=5).

Equation 4

S(a-1，b/2)≥S(a，b)+S(a，b+1)

A=0 wherein ... 5, b=0 ... B-1,

" a-1 " corresponding to the layer of i piece, " a " is corresponding to the layer of i+1 piece.

To step S280, but this is not construed as limiting the present invention by following C class false code 2 performing step S210.

False code 1

for(a＝0；a＜＝5；a++){ //for?all?levels

pbuf＝buf[0][0]； //pointer?to?target?buffer

B＝1＜＜a； //block?length?in?levela

for(b＝0；b＜B；b++){ //forall?blocks

if(lskip[a][b]) //if?block?can?not?be?skipped

size[a][b]＝EncodeBlock(x+b＊B，buf[a][b])；//encode?block?and?store?in?buf

}

if(a＞0){ //if?not?highest?level

for(b＝0；b＜B；b+＝2){

if(lskip[a][b]){//compare?size?of?two?current?blocks?with?size?of?one?block?in?level?a-1

if(size[a-1][b/2]＞size[a][b]+size[a][b+1]){//copy?two?short?blocks?from?current?level?a?into

target?buffer

memcpy(pbuf，buf[a][b]，size[a][b])；

memcpy(pbuf+size[a][b]，buf[a][+1]，size[a][b+1])；

pbuf+＝size[a][b]+size[a][b+1]；//increment?target?buffer

}

else{

pbuf+＝size[a-1][b/2]；//increment?target?buffer

//all?subordinate?shorter?blocks?in?lower?levels?can?be?skipped

for(aa＝a+1；aa＜＝5；aa++)//for?all?lower?levels

for(bb＝(aa-a)＊2＊b；bb＜(aa-a)＊2＊(b+1)；b++)//for?all?subordinate?blocks

skip[aa][bb]＝1；//set?skipping?flag

}

}

else

pbuf+＝GetSkippedSize0；//increment?target?buffer(add?size?of?skipped

blocks)

}

}

}

Figure 11 is the example flow diagram of the block switching method that is used for audio signal of version according to another embodiment of the present invention, and Figure 12 is the exemplary plot that is used to explain Figure 11 theory.Specifically, to end method, this method only stops when piece improves to two layers rather than to a layer version of another embodiment corresponding to the top of expanding certainly.This is to push up the key distinction to end method with the front certainly with reference to what Figure 10 described, and if block is only progressive to a layer, and the method for Figure 10 will stop.

Referring to Fig. 4 and Figure 11, initial comparing unit 110b, as step S210, will (at top layer) first position size compare (S310) with a position size of second.No matter the comparative result of step S310 how, initial comparing unit 110b compares (S320 and S370) with second position size and the 3rd position size.If first position size is less than second position size (being judged to be "No" among the S310), and second position size is less than position size (being judged to be "No" among the S320) (referring to " situation E " among Figure 12 and " situation F ") of two the 3rd, promptly second of first ratio and the 3rd are more useful, then initial comparing unit 110b selects first as optimical block (S330), and stop the comparison (, will note pentagram especially) of one deck down referring to " the situation F " among Figure 12.Otherwise, if second position size is equal to or greater than the 3rd position size (being judged to be "Yes" among the step S320), then initial comparing unit 110b based on first and and the 3rd comparative result determine to select first still to be to compare at one deck down.Specifically, if first the 3rd of ratio more useful (being judged to be "No" among the step S340), then initial comparing unit 110b selects first (S350) (" the situation E " referring among Figure 12 will note pentagram especially).Otherwise (S340 is judged to be "Yes" in step), condition comparing unit 110c is with the 3rd and the 4th comparison, and, select best (S360) (referring to " situation D " Figure 12) from the 3rd, the 4th and the 5th then with the 4th and the 5th comparison.

Simultaneously, if second position size is equal to or greater than two the 3rd position size (being judged to be "Yes" among the step S320), and first position size is equal to or greater than second position size (being judged to be "Yes" among the step S310), if and second position size is less than the 3rd (being judged to be "No" among the step S370) (referring to " case B " among Figure 12 and " situation C "), then condition comparing unit 110c is temporary transient selects second (referring to four jiaos of stars in " case B " and " situation C "), and compares (S380) at following one deck.Otherwise, the 3rd less than first and second (being judged to be "Yes" among the S370) (referring to " the situation A " among Figure 12), condition comparing unit 110c is temporary transient to select the 3rd (referring to four jiaos of stars in " situation A "), and the 3rd is compared with the 4th and the 4th and the 5th.

Long-term forecasting (LTP)

Most of sound signals have the harmonic wave or the periodic component of basic frequency of being derived from or music instrument keynote.Because require very high time item, short-term forward direction adaptive prediction device is difficult to remove these the distance sample correlativity, thereby causes unfavorable secondary quantity of information.In order more effectively to utilize the correlativity that has between the distance sample, can carry out long-term forecasting.

Figure 13 is the example block diagram according to the long-term forecasting device that is used for audio signal of the embodiment of the invention, and Figure 14 is the exemplary process diagram according to the long-range forecast method that is used for audio signal of the embodiment of the invention.Referring to Figure 13, long-term predictor 190 comprises that lag information determines parts 190a, filter information estimation section 190b and decision component 190c, and long-term predictor 190 uses the short-term residual value e (n) of input to produce long-term forecasting value e^ (n).In brief, can calculate long-term forecasting value e^ (n) and residual value e～(n) for a long time according to following equation 5, but this is not construed as limiting to the present invention.

Equation 5

$\tilde{e}\left(n\right)=e\left(n\right)-\hat{e}\left(n\right)=e\left(n\right)-\underset{j=-2}{\overset{2}{\mathrm{\Σ}}}{\mathrm{\γ}}_j\·e(n-\mathrm{\τ}-j)$

Wherein τ represents that sample lags behind γ _jThe LTP filter coefficient of expression through quantizing, the new residual value after the expression long-term forecasting of and e～(n).In conjunction with Figure 13, Figure 14 the long-term forecasting processing is described.

Referring to Figure 13 and Figure 14, long-term predictor 190 is skipped the normalization (S410) of following input signal.

Equation 6

$e_{\mathrm{norm}}\left(n\right)=e\left(n\right)\·\frac{\sqrt{\left|e\left(n\right)\right|}}{1+5\·\sqrt{\stackrel{\‾}{\left|e\left(n\right)\right|}}},$

Wherein | e (n) | be the arithmetical mean of absolute value.If save the normalization of input value, then can lower the long-term forecasting complexity.Yet if adopt random access, normalization is still useful, to avoid non-optimal compression.

Then, lag information determines that parts 190a uses autocorrelation function to determine lag information τ (S420).Use equation 7 to calculate autocorrelation function (ACF).

Equation 7

$r_{\mathrm{ee}}\left(\mathrm{\τ}\right)=\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}e\left(n\right)\·e(n-\mathrm{\τ}),\mathrm{for\τ}=K+1...K+{\mathrm{\τ}}_{\max }$

Wherein K is a short-term forecasting time item, and Δ τ _MaxBe that maximum relatively lags behind, Δ τ wherein _Max=256 (for example for the 48kHz audio materials), 512 (for example 96kHz) or 1024 (for example 192kHz), this depends on sampling rate.At last, use maximum absolute ACF value max|r _Ee(τ) | position is as the best τ that lags behind.In addition, can use the quick A CF algorithm of using FFT (fast fourier transform) to replace direct ACF to calculate.If the frequency domain at similar FFT is carried out the ACF algorithm, then can reduce scramble time and complexity.

Then, filter information estimation section 190b uses Wiener-Hopf equation estimation filter information γ based on stable state _j(S430).The unstable state form of Wiener-Hopf equation is an equation 8.

Equation 8

$\left[\begin{array}{c}r(\mathrm{\τ}-\mathrm{2,0})\\ r(\mathrm{\τ}-\mathrm{1,0})\\ r(\mathrm{\τ},0)\\ r(\mathrm{\τ}+\mathrm{1,0})\\ r(\mathrm{\τ}+\mathrm{2,0})\end{array}\right]=\left[\begin{array}{ccccc}r(\mathrm{\τ}-2,\mathrm{\τ}-2)& r(\mathrm{\τ}-2,\mathrm{\τ}-1)& r(\mathrm{\τ}-2,\mathrm{\τ})& r(\mathrm{\τ}-2,\mathrm{\τ}+1)& r(\mathrm{\τ}-2,\mathrm{\τ}+2)\\ r(\mathrm{\τ}-1,\mathrm{\τ}-2)& r(\mathrm{\τ}-1,\mathrm{\τ}-1)& r(\mathrm{\τ}-1,\mathrm{\τ})& r(\mathrm{\τ}-1,\mathrm{\τ}+1)& r(\mathrm{\τ}-1,\mathrm{\τ}+2)\\ r(\mathrm{\τ},\mathrm{\τ}-2)& r(\mathrm{\τ},\mathrm{\τ}-1)& r(\mathrm{\τ},\mathrm{\τ})& r(\mathrm{\τ},\mathrm{\τ}+1)& r(\mathrm{\τ}+1,\mathrm{\τ}+2)\\ r(\mathrm{\τ}+1,\mathrm{\τ}-2)& r(\mathrm{\τ}+1,\mathrm{\τ}-1)& r(\mathrm{\τ}+1,\mathrm{\τ})& r(\mathrm{\τ}+1,\mathrm{\τ}+1)& r(\mathrm{\τ}+1,\mathrm{\τ}+2)\\ r(\mathrm{\τ}+2,\mathrm{\τ}-2)& r(\mathrm{\τ}+2,\mathrm{\τ}-1)& r(\mathrm{\τ}+2,\mathrm{\τ})& r(\mathrm{\τ}+2,\mathrm{\τ}+1)& r(\mathrm{\τ}+2,\mathrm{\τ}+2)\end{array}\right]\·\left[\begin{array}{c}{\mathrm{\γ}}_{-2}\\ {\mathrm{\γ}}_{-1}\\ {\mathrm{\γ}}_0\\ {\mathrm{\γ}}_1\\ {\mathrm{\γ}}_2\end{array}\right]$

Therefore, calculate for j k=-2 ... 2 ACF value r _Ee(τ+j, 0) and r _Ee(τ+j, τ+k).Because this matrix is symmetrical, therefore only calculate top-right triangle (15 values).Yet, because therefore hypothesis unstable state form can't reuse the stable state r that calculates at the best hysteresis searching period _Ee(τ) value.

Simultaneously, if stable state, promptly r (j, k)=r (j-k), then therefore can use the form stable of Wiener-Hopf:

Equation 9

$\left[\begin{array}{c}r(\mathrm{\τ}-2)\\ r(\mathrm{\τ}-1)\\ r\left(\mathrm{\τ}\right)\\ r(\mathrm{\τ}+1)\\ r(\mathrm{\τ}+2)\end{array}\right]=\left[\begin{array}{ccccc}r\left(0\right)& r\left(1\right)& r\left(2\right)& r\left(3\right)& r\left(4\right)\\ r\left(1\right)& r\left(0\right)& r\left(1\right)& r\left(2\right)& r\left(3\right)\\ r\left(2\right)& r\left(1\right)& r\left(0\right)& r\left(1\right)& r\left(2\right)\\ r\left(3\right)& r\left(2\right)& r\left(1\right)& r\left(0\right)& r\left(1\right)\\ r\left(4\right)& r\left(3\right)& r\left(2\right)& r\left(1\right)& r\left(0\right)\end{array}\right]\·\left[\begin{array}{c}{\mathrm{\γ}}_{-2}\\ {\mathrm{\γ}}_{-1}\\ {\mathrm{\γ}}_0\\ {\mathrm{\γ}}_1\\ {\mathrm{\γ}}_2\end{array}\right]$

If direct ACF is used for determining best the hysteresis, then only calculates r _Ee(K+1 ... K+ τ _Max).Comparatively speaking, always be to use the quick A CF of FFT to calculate r _Ee(0 ... N-1).Therefore, not necessarily to calculate the value r (0 that in stable state Wiener-Hopf equation, requires once more ... 4) and r (τ-2 ... τ+2), obtain among the result of the quick A CF that from step S420, the search that lags behind has been finished but simply.

Decision component 190c uses the filter information γ that estimation obtains among lag information τ definite among the step S420 and the step S430 _jProduce long-term forecasting value e^ (n).

Then, decision component 190c calculates the bit rate (S450) of sound signal before to audio-frequency signal coding.In other words, decision component 190c calculates short-term residual value e (n) and the in fact uncoded bit rate of residual value e～(n) for a long time.Specifically, under the situation of calculating the Rice encoded bit rate, decision component 190c can determine the optimum code parameter of residual value e (n), e～(n) by function G etRicePara (), and calculate necessary figure place by function G etRiceBits (), with residual value e (n), the e～(n) encode to defining by code parameter, but this is not construed as limiting the present invention.

Decision component 190c judges long-term forecasting whether useful (S460) based on the bit rate of calculating among the step S450.According to the judgement of step S460,, then do not carry out long-term forecasting, and make procedure termination if long-term forecasting is unhelpful (being judged to be "No" among the step S460).On the contrary, if long-term forecasting is useful (being judged to be "Yes" among the step S460), then decision component 190c determines to use long-term forecasting and export long-term forecasting value (S470).In addition, decision component 190c can be with lag information τ and filter information γ _jBe encoded to secondary information, and set the flag information that indicates whether to carry out long-term forecasting.

It will be understood by those skilled in the art that and to make multiple correction and variation to the present invention and do not break away from the spirit or scope of the present invention.Therefore, the present invention is intended to cover all corrections and the variation of invention, supposes that they drop in the scope of appended claims and equivalent thereof.

Commercial Application

Therefore, the present invention is applicable to that audio frequency can't harm (ALS) Code And Decode.

Claims (34)

1. method that is used for audio signal comprises:

Received audio signal; And

Handle the sound signal that is received;

Wherein handle described sound signal according to scheme, described scheme comprises: the size information of at least two A+1 layer pieces and size information corresponding to the A layer piece of described at least two A+1 layer pieces are compared; And

If the size information of described at least two A+1 layer pieces less than the size information of described A layer piece, then is defined as optimical block with described at least two A+1 layer pieces.

2. the method for claim 1 is characterized in that, described size information corresponding to coding result, position size and coded data piece one of them.

3. the method for claim 1 is characterized in that, described A layer piece is corresponding to the combination of at least two A+1 layer pieces.

4. method as claimed in claim 3 is characterized in that hierarchy has at least two layers, and top block length is corresponding to the integral multiple of the block length of lowermost layer.

5. method as claimed in claim 4 is characterized in that, described hierarchy has six layers, and described top block length is corresponding to 32 times of the block length of lowermost layer.

6. the method for claim 1 is characterized in that, the size information of at least two A+1 layer pieces is corresponding to the size of an A+1 layer piece and the big or small sum of next A+1 layer piece.

7. the method for claim 1 is characterized in that, also comprises:

If the size information of described at least two A+1 layer pieces is greater than the size information of described A layer piece, then the size information of A layer piece and the size information of A-1 layer piece compare at least.

8. method as claimed in claim 7 is characterized in that, also comprises:

If the size information of described at least two A layer pieces less than the size information of A-1 layer piece, then is defined as optimical block with described at least two A layer pieces.

9. the method for claim 1 is characterized in that, receives described sound signal as broadcast singal.

10. the method for claim 1 is characterized in that, also comprises:

On digital media, receive described sound signal.

11. a method that is used for audio signal comprises:

Received audio signal; And

Handle the sound signal that is received,

Wherein handle described sound signal according to scheme, described scheme comprises:

The size information of A layer piece and the size information of at least two A+1 layer pieces are compared; And

If the size information of described A layer piece less than the size information of described at least two A+1 layer pieces, then is defined as optimical block with described A layer piece.

12. method as claimed in claim 11 is characterized in that, described A layer piece is corresponding to the combination of at least two A+1 layer pieces.

13. method as claimed in claim 11 is characterized in that, receives described sound signal as broadcast singal.

14. method as claimed in claim 11 is characterized in that, also comprises:

On digital media, receive described sound signal.

15. a method that is used for audio signal comprises:

Received audio signal; And

Handle the sound signal that is received;

Wherein handle described sound signal according to scheme, described scheme comprises:

The size information of A layer piece and the size information of at least two A+1 layer pieces are compared;

The size information of A+1 layer piece and the size information of at least two A+2 layer pieces are compared; And

If the size information of described A layer piece then is defined as optimical block with described A layer piece less than the size information of described at least two A+1 layer pieces and the size information of described at least four A+2 layer pieces.

16. a method that is used for audio signal comprises:

Received audio signal;

Handle the sound signal that is received;

Wherein handle described sound signal according to scheme, described scheme comprises:

On the entire frame of described sound signal, the size information of A layer piece and size information corresponding at least two A+1 layer pieces of described A layer piece are compared; And

If all size information of included A layer piece all less than the size information of described at least two the A+1 layer piece corresponding with described A layer piece, then are defined as optimical block with A layer piece in the described frame.

17. a computer-readable medium that stores instruction on it, described instruction makes the processor executable operations, and described operation comprises:

The size information of at least two A+1 layer pieces and size information corresponding to the A layer piece of described at least two A+1 layer pieces are compared; And

If the size information of described at least two A+1 layer pieces less than the size information of described A layer piece, then is defined as optimical block with described at least two A+1 layer pieces.

18. a computer-readable medium that stores instruction on it, described instruction makes the processor executable operations, and described operation comprises:

The size information of A layer piece and the size information of at least two A+1 layer pieces are compared; And

If the size information of described A layer piece less than the size information of described at least two A+1 layer pieces, then is defined as optimical block with described A layer piece.

19. a device that is used for audio signal comprises:

Initial comparing unit, described initial comparing unit will at least two A+1 layer pieces size information and compare corresponding to the size information of the A layer piece of described at least two A+1 layer pieces; And

The condition comparing unit, if the size information of described at least two A+1 layer pieces less than the size information of described A layer piece, then described condition comparing unit is defined as optimical block with described at least two A+1 layer pieces.

20. a device that is used for audio signal comprises:

Initial comparing unit, described initial comparing unit compares the size information of A layer piece and the size information of at least two A+1 layer pieces; And

The condition comparing unit, if the size information of described A layer piece less than the size information of described at least two A+1 layer pieces, then described condition comparing unit is defined as optimical block with described A layer piece.

21. a method that is used for audio signal comprises:

The size information of at least two A+1 layer pieces and size information corresponding to the A layer piece of described at least two A+1 layer pieces are compared; And

If the size information of described at least two A+1 layer pieces less than the size information of described A layer piece, then is defined as optimical block with described at least two A+1 layer pieces.

22. a method that is used for audio signal comprises:

The size information of A layer piece and the size information of at least two A+1 layer pieces are compared; And

If the size information of described A layer piece less than the size information of described at least two A+1 layer pieces, then is defined as optimical block with described A layer piece.

23. a method that is used for audio signal comprises:

Received audio signal;

Handle the sound signal that is received;

Wherein handle described sound signal according to scheme, described scheme comprises:

The size information of at least two A+1 layer pieces and size information corresponding to the A layer piece of described at least two A+1 layer pieces are compared;

If the size information of described at least two A+1 layer pieces less than the size information of described A layer piece, then is defined as optimical block with described at least two A+1 layer pieces;

Autocorrelation function based on the sound signal that comprises described optimical block is determined lag information; And

Estimate the long-term forecasting filter information based on described lag information.

24. method as claimed in claim 23 is characterized in that, also comprises:

Before to described coding audio signal, estimate the bit rate of described sound signal.

25. method as claimed in claim 24 is characterized in that, also comprises:

Based on estimated bit rate described lag information and described long-term forecasting filter information are encoded to secondary information.

26. method as claimed in claim 23 is characterized in that, also comprises:

Calculate the autocorrelation function of described sound signal in frequency domain.

27. method as claimed in claim 23 is characterized in that, estimates the operation of long-term forecasting filter information based on stable state.

28. method as claimed in claim 27 is characterized in that, uses autocorrelation function to carry out the operation of estimating the long-term forecasting filter information.

29. method as claimed in claim 23 is characterized in that, the sound signal of described sound signal before corresponding to normalization.

30. method as claimed in claim 23 is characterized in that, described sound signal receives as broadcast singal.

31. method as claimed in claim 23 is characterized in that, also comprises:

On digital media, receive described sound signal.

32. a computer-readable medium that stores instruction on it, described instruction makes the processor executable operations, and described operation comprises:

The size information of at least two A+1 layer pieces and size information corresponding to the A layer piece of described at least two A+1 layer pieces are compared;

If the size information of described at least two A+1 layer pieces less than the size information of described A layer piece, then is defined as optimical block with described at least two A+1 layer pieces;

Autocorrelation function based on the sound signal that comprises optimical block is determined lag information; And

Estimate the long-term forecasting filter information based on described lag information.

33. a device that is used for audio signal comprises:

Initial comparing unit, described initial comparing unit will at least two A+1 layer pieces size information and compare corresponding to the size information of the A layer piece of described at least two A+1 layer pieces;

The condition comparing unit, if the size information of described at least two A+1 layer pieces less than the size information of described A layer piece, then described condition comparing unit is defined as optimical block with described at least two A+1 layer pieces;

Lag information is determined parts, and described lag information determines that parts determine lag information based on the autocorrelation function of the sound signal that comprises optimical block; And

The filter information estimation section, described filter information estimation section is estimated the long-term forecasting filter information based on described lag information.

34. a method that is used for audio signal comprises:

The size information of at least two A+1 layer pieces and size information corresponding to the A layer piece of described at least two A+1 layer pieces are compared;

If the size information of described at least two A+1 layer pieces less than the size information of described A layer piece, then is defined as optimical block with described at least two A+1 layer pieces;

Autocorrelation function based on the sound signal that comprises optimical block is determined lag information; And

Estimate the long-term forecasting filter information based on described lag information.

Images